# Sandal > Infrastructure for agents. Sandal gives every AI agent somewhere safe to run: Firecracker/Cloud Hypervisor microVM sandboxes, GitHub Actions runners (Linux/Windows/macOS), iOS/Android simulators, and a code interpreter — one SDK, bare-metal economics. Sandal runs on dedicated Hetzner/OVH/Leaseweb hardware and Mac minis instead of hyperscaler markup, so prices are far lower while every sandbox still gets its own kernel. Per-second billing, paused workloads cost $0, warm claims target <200ms, cold creates target <1s. EU and US regions; SOC2 in progress. ## Site - [Homepage](https://sandal.dev/): Product overview, pricing teaser, terminal quickstart (`npx sandal init`), and logos of teams building on Sandal. ## Docs - [Docs home](https://sandal.dev/docs): Documentation index and navigation. - [Quickstart](https://sandal.dev/docs/quickstart): Install an SDK, create your first sandbox, and run a command in under five minutes. ### Products - [Sandboxes](https://sandal.dev/docs/sandboxes): Firecracker/Cloud Hypervisor microVMs — create, exec, fs read/write, expose ports, pause/resume, fork. - [Runners](https://sandal.dev/docs/runners): Drop-in GitHub Actions runners via a `runs-on:` label swap; JIT-registered Linux/Windows/macOS pools. - [Simulators](https://sandal.dev/docs/simulators): iOS and Android simulators with 60fps streaming, Appium/ADB tunnels, and a React `` component. - [Xcode builds](https://sandal.dev/docs/xcode-builds): Remote Xcode builds on dedicated M4 hosts with content-addressed source sync, signing grants, and cached DerivedData. - [Code interpreter](https://sandal.dev/docs/code-interpreter): Sandboxed code execution for agent-generated code with per-second billing. ### Persistence - [Snapshots and forks](https://sandal.dev/docs/snapshots-and-forks): Per-minute incremental encrypted snapshots (V4/RAFS), lazy restore with prefetch playbooks, and instant forking. ### Reference - [TypeScript SDK](https://sandal.dev/docs/sdk-typescript): `@sandal/sdk` — full API reference and examples. - [Python SDK](https://sandal.dev/docs/sdk-python): `sandal` package — full API reference and examples. - [CLI](https://sandal.dev/docs/cli): The `sandal` command-line tool for scripting and local development. - [API reference](https://sandal.dev/docs/api-reference): Raw REST API endpoints underlying every SDK. - [MCP server](https://sandal.dev/docs/mcp): Expose Sandal sandboxes, runners, and simulators to MCP-compatible agents. - [Webhooks](https://sandal.dev/docs/webhooks): Event notifications for sandbox lifecycle, runner jobs, and simulator sessions. ### Platform - [Architecture](https://sandal.dev/docs/architecture): How Firecracker/Cloud Hypervisor microVMs, metering, and regions fit together. - [Security](https://sandal.dev/docs/security): Isolation model, per-instance DPoP-scoped tokens, and SOC2 status. - [Pricing](https://sandal.dev/docs/pricing): Active-CPU + provisioned-RAM metering, per-second billing, runner and simulator rates. - [Limits](https://sandal.dev/docs/limits): Rate limits, quotas, and resource ceilings per plan. - [Comparison](https://sandal.dev/docs/comparison): Evidence-gated capability matrix against alternative sandbox/runner providers — every number generated from qualification runs. ## Optional - [Every docs page as Markdown](https://sandal.dev/docs): Each page above mirrors at the same path with a `.md` suffix (e.g. `https://sandal.dev/docs/quickstart.md`) for direct LLM ingestion.